Detection is futile...
The idea that we can rely on AI systems to detect deepfakes seems to ignore the fact that the very reason GANs work so well is because they are trained against an "adversarial" network until it can't spot the difference. All we'll achieve is ever increasingly perfect deep fakes that is practically impossible to detect.
It also seems ironic that the go-to application of any new on-line technology - porn - is probably training our future overlords. At least they'll know how to have a good time.
Hmmm...
Jim